Via Bell tests, in which the two particles, in each entangled pair, are simultaneously measured at different, random (polarization or spin) angles. The correlation statistics between these pairs, are then computed, after the fact. The December 2018 issue of Scientific American has a good article describing recent experiments, along with the standard (incorrect) interpretation of why the correlations occur.
The Matlab code in my vixra paper performs exactly this type of test on 1,000,000 polarized coins, and demonstrates that, contrary to popular belief, the correlations can be reproduced classically, and have nothing to do with either spooky action at a distance, or hidden variables, which are usually assumed to be the ONLY two possible explanations for the correlations.
I sent a Letter to the Editors of Scientific American, about this, over two months ago. But I doubt that it will ever be published - too controversial.